Doesn't take into account future price rise of BTC. If they are worth $500 in a years time then is your hardware still profitable? By the way $0.087 /kWh is low, it's $0.28+ where I live...

A loss is a loss, in this case the magnitude of one is greater than the other...

There are some altcoins that are still profitable mining solo. More so than mining BTC. It depends a bit on your experience and hardware knowledge too, to get the most out of your system. I don't think many people will write this guy a how-to guide given there are so many resources like the one you linked, and people mining altcoins dont really want a influx of new miners with GPU wrecking 'their' altcoin's difficulty.

Take 100 (one hundred) Radeon 7990's. Overclock cores. Overclock memory. Increase voltages.

Now mine scrypt coins for 1 week on all of them.

Record number of cables on fire.

SCIENCE!

Sometimes cards can limp along on just a few of the phases of the VRMs and appear to work normally when not under load. It's fairly well known that phases can fail (usually power transistor failures from overtemp on dual GPU cards) but the GPU will still 'sort of work'. As soon as you give it load it cant supply the correct amperage and power stability to the GPU so it will crash or lock up.

http://www.geeks3d.com/20100504/tutorial-graphics-cards-voltage-regulator-modules-vrm-explained/

Ideally put it in a different computer for testing. Cards can work in 2d modes and video modes but still be damaged enough to crash at higher loads (eg gaming, stress testing, mining).

Usually the leads aren't rated for it. I guess that's a quality multimeter for real mains testing then (ie. not <$100 hobbyist type). I've melted cheap ones with as little as 100 watts.

Wait, are you measuring 1200+ watts through your multimeter? 

So far there have been no leaked drops. It will be really easy to tell if/when it happens because it will dry into a trail of borax which will appear as a white powder on the cups. Given that  I've obviously gone for function over form I think it's a pretty reasonable precaution to take. I put the side of the case on when I have visitors around so it's hidden from view. The alternative was to reposition the case on its side and upside down (eg rotated 90 degrees anticlockwise) so that the drops would fall downwards away from the boards but this makes accessing and checking on things very difficult.

When I first setup the 5970 with watercooling I had a glycol coolant loop that leaked onto the bottom of the case (there was no 5750 at the time) and some of it dried into a horrible green mass of glycol coolant. That turned out to be an epic fail because a lot of microorganisms thrive on glycol as it turns out, so I decided to use borax as an experimental alternative. So far I am happy with it. In the future all of my work with watercooling will most likely be done with silicone-type tubing (think medical tubing) because it forms a far superior fit on any connectors that are used and has a much lower risk of leaking. The chemical stability of silicones versus PVC is also another advantage.

It wastes a lot of my time and money. It's like a consolation prize.

Yes it will use less power. The answer why is slightly complicated. It depends on how many power transistors are present in the VRMs of the card/card(s). Because the dual GPU cards do more with less power transistors, there is less power lost as heat in these components and so you end up with a more efficient system. Eg in dual GPU cards the ratio of power used by the cores:VRMs is higher than in two equivalent single GPU cards.

Ok I'll bite. Here's my mining rig:



Wait... we'll have to look deeper.

---

Here's the case: From top to bottom, we have: Radeons 6870, 5970, and 5750. The top two are watercooled and the 5750 is just a piece of crap thrown in for lulz. Note the wonderful use of garden-rated PVC tubing+connectors on the 5970 and the patented Bubble-Loop(TM) that prevents all the liquid draining out of the 6870 waterblock. Despite having two watercooled cards they are not amazingly overclocked or anything, it was mostly to keep them quiet and cool.

---

What is the crazy blue thing bottom left? Well its the two reservoirs ($4 Ikea storage containers) for each cooling loop. There is a total of about 4 liters (including tubing volume: substantial) in the top reservior (for the 5970) and 3 liters in the bottom reserviour (6870). The cooling fluid is distilled water saturated with borax (sodium tetraborate) which has been filtered to remove sodium oxide. I won't explain why I've done this. Suffice to say that nothing biological known to science can grow in this solution, it's fairly stable and harmless from a chemical perspective and aluminum is stable at it's pH which is roughly 9.7. The solution is colourless and transparent when prepared. (There is about 1 kg of borax solid dissolved to make the solution for both reservoirs). It has a 1 watt LED array that was like $0.80 on ebay to light it up. The little box I made that sits on top is an on/off switch for the light and also one of the pumps.

So why are they blue if borax solution is colorless to start with? Well the 5970's cooling loop has two radiators that I use. One is for winter (shown below) and the other is for summer, which is installed outside and runs through some tubing installed through the floor. The outside radiator is substantial in size, and has 8 meters of internal tubing length alone and several m^2  cooling fin surface area. The inside radiator is smaller but still probably overkill. It is 40x30x3 cm in size roughly. The total tubing length in winter for the 5970 loop is ~9 meters (15 in summer) and the 6870 loop is ~10 meters. The flow rates of both are less than 1 liter per minute (very low in terms of PC watercooling, apparently). The blue colour comes from copper ions leaching out of the brass or copper that the radiator(s) are made out of (the tubing is copper or brass but the fins are aluminum). The temps are pretty good on the cards, the cores of the 5970 are currently (25 deg C ambient) at 41 deg C with a slight overclock, more importantly VRMs sit <80 deg C. The 6870 is at 47 deg C with the pump spinning so slow it is silent, it drops to high 30's when the pump runs at its normal speed.

Despite the relatively high electrical conductivity of the solution, because of the stupidly long tubing distance between the waterblocks and the radiators (which are dissimilar metals in each case, subject to electrochemical processes ie. corrosion) I have measured the conductance between them and only nano-ampere currents can pass between the two (which is good!).

---

One of the pumps (which I fully expect to drop a hose and leak badly one day) and masses of tubing running behind the HTPC.

---

In winter the radiator used to cool the 5970 loop is installed on the back of the air intake of a portable conditioner that runs 24/7 in fan mode. It uses about 30 watts. Its good to move air around the room too, dispersing the warmth from the 5970 You'll marvel at my amazing placement of coolant lines carrying highly conductive solution right above several bunches of 240v cables, connectors, powerboards etc. However I have confident that these connections won't leak because it has ran this was for more than 2 seasons now.
In the corner you can see three tubes running through the wall/floor, two are the coolant lines for the summer radiator and the third is just the air conditioners water drain. The black/red cables beside it are to power the fans on the summer radiator.

---

The 6870 cooling loop is a much more recent addition and I setup this extremely dodgy looking radiator to leak test the system and do some other performance tests. I made some crazy electronics to control the pump speed by toggling between 5v and 12v rails from the PSU but eventually found it was not sufficient for my needs, so instead the $8 pump now runs of $200 lab bench power supply until I can be bothered fixing it up 'properly'. The OUTRAGEOUS amber mass next to the radiator (which has 4x 120mm fans ontop) is because of an absolutely ridiculous air-bubble-sucking problem I encountered where the pump was pulling such an extreme vacuum on one half of the cooling loop that it was pulling air between the tubing-connector join near the radiator. One hour with a hot glue gun and several zipties putting kilograms of force on it didn't solve it. The solution was to move the pump position in the loop and lower its voltage a little.

For now it tentatively lives on the glass shelf under my TV.

---

Last thing perhaps worth mentioning with all this highly electrically conductive coolant around the place and dodgy agricultural grade tubing/connectors being used... is the amazing drip-mitigation feature I have installed to protect the lower-lying cards in the case. As you might be able to see, it is in fact a collection of disposable drinking cups cut up and taped together to redirect any liquid to the bottom and out of the case! $2 of plastic may yet save the $100+ cards! Wheee!

To make it worse, PSU conversion efficiency does not vary linearly with load.

---

OP Should look at underclocking memory speeds to save on power too (and reduce card temps), assuming he's mining SHA-256 coins. (Don't underclock if running scrypt, usually).

Been through dozens of similar units. You'll get over the noise and heat from them. Then the issues of inconvenient OS that are required... the specs might seem amusing at first but it will wear off. You could've probably picked up 2 or 3 of these at a government surplus auction for that price, they sometimes have so many that there arn't enough auction goers to take them home, so the auctioneer more or less gives them away for free or forced into other auctioned items so that they clear the floor for the day.

Pro tip: You'll make far more money (over time) selling them as parts than you will selling them as whole. $180? That's less than the value of 4x SAS drives of *any* size alone.

Well have you read about the projected difficulty increases over the next 12 months and run some calculations with a few profitability calculators? You'll probably be very dissapointed. You'll probably not make back your initial investment (at least 0.89 BTC/ea at present).

On the other hand, if you want it for non-profit purposes then go for it.

Have you looked up the power draw on those cards and then estimated your PSU requirements for the rest of the PC with one of the PSU calculators? How does it compare to the PSU you are running? I'd recommend at least 20% extra (as a minimum, given the load you are running it at) to accommodate for lowering power output capability in the months/years to come. 20% should give you room for a year, if you are lucky, on whatever PSU you find you need to get to make the setup stable.

You could consider running two PSUs on the one computer. I'm sure you're not the first person to setup a rig like that around here.

You'd also do well to get a 92% conversion efficiency PSU next time, not a ~80%. If you work out the saving from that 12% at 1000 watts or more running 24/7/356... it's a lot of money.

Work out your total Mh/s (hopefully many Gh/s if you intend to solo mine...) and use one of the many profitability calculators to work out how long it will take you on average to find a block (it will almost certainly be many months or years) and if this is acceptable to you then you can solo mine. But realistically you're better off with a pool as Voodah says.

This site is probably a good place to ask your questions or get info. You'll need more posts / 4 hours logged in to post in the other sub-forums though.

88 deg C is getting up there. Have you checked that the other temps in your card is not much higher than that? Most 6xxx and 7xxx cards have 3 thermocouples per GPU, sometimes the one reported in cgminer is the lowest temp of the three.

Hi there good to see some members with some suggestions to make, seems like a lot of the newbie forum is scammers and spammers or those desperate to make a quick buck.